Automatic parameter regulation for CT/MRA viscous fluid registration

In medical image processing, automatic parameter regulation reduces labor costs of the visually checking that the organ contours of CT and MRA image are overlapped well or not. The present paper compares three similarity measures of the mutual information, the gradient difference, and the gradient image mutual information for automatic parameter regulation. The parameter regulation chooses one of the ten kinds of numerical value. We define a best parameter, some good parameters that have a gap of CT and MRA organ contours within 10mm and the rest poor parameters over 10mm to the 10 different value of a parameter for a MRA image. We measure the best parameter detection rate, good parameter detection rate and the number of checking till the good parameter detected. Experimental results reveal that the gradient image mutual information provides the best result, the best parameter detection rate at first is 24%, the good parameter detection rate at first is 63%, and the number of confirmations to regulate the good parameter is 220 for 109 regulating images.